Beyond basic prompting, what is your experience with building applications that integrate with LLM APIs? Please describe any projects where you've used advanced techniques like Retrieval Augmented Generation (RAG), agentic workflows, or tool calling.